Preventing crashes, designing words

nina's picture

Are there any fonts that use contextual alternates with chopped-off serifs, like the ones below, to prevent crashes and/or enhance "word images", and do you think it makes sense?

I'm doing something similar to my f + ascender combinations, but that seems more conventional than the r + x-line-serif combinations, at least I've seen it done before (and also discussed it with a couple of people), so I'm less uncertain about those.

I'm pretty excited about these, but would love some second eyes/opinions before I take this too far (it obviously is a pretty expandable concept :-) ).

Dunno – are these too "designy"? (Maybe specifically the "rt"?)

Oh, and please excuse the still somewhat wobbly spacing, I'm in the middle of fixing it.

guifa's picture

It looks great Nina. I've also been experimenting with these discreet contextual alternatives. Perhaps instead of (or in addition to) chopping off serifs, you could also slightly alter the r, starting the arm a bit lower so as to make a higher angle (and thus a skinnier glyph).

There's no taking it too far. Now if Eben would publish a nice compilation of all of his research for us to gobble up .... *winkwink* *nudgenudge* *stabstab*

dan_reynolds's picture

Nina, in text printed at text sizes, you won't see the difference between column 1 and column 2, I think. Especially in the rt combination. Maybe something for a display variant…

nina's picture

Oh right Matthew, now I remember you started a thread on something similar recently. Was it the descender of the "j" in the "gj" combo? Or something?
I'm not sure I'd want to make the "r" skinnier – at least not a lot, as I feel it needs a certain width in this wide font of mine – which is why I started experimenting with "the other side"…


That's 9pt, which is the target size of the font. I admit the "rt" is not too noticeable (I guess that gap's too small), but the others seem to do some more or less subtle de-cluttering/de-blob-making to my eye.

eliason's picture

(With the first three) I get a sense of overlapping (a third dimension), which may be no less distracting than a crash.

Maybe with those, a shortened serif (so there'd still be a notch underneath) would be better than the diagonal shear.

eliason's picture

Lateral thinking idea: instead of trying to pull apart the top, why not try to push the bottom together: for example, when followed by a v/w/t, have r's bottom right serif extend further.

dezcom's picture

Nina, I use {calt} on some of my fonts for better fit and preventing crashes. Also, Eben gave a presentation on this at TypeCon and ATypI last year.

By all means try it!


nina's picture

Aw. The 3D thing probably isn't good.

To give another reason for why I'd like to prevent those crashes:
In German typesetting, as is commonly discussed with regard to f-ligatures, connecting ligatures are forbidden across more than one part of a compounded word. Thus, say in the word aufleben (to revive) –which is composed of the prefix auf and the verb leben (to live)–, since it's verboten to set that with an "fl" ligature, but the crash looks bad too, most typesetters will then track that "fl" combination out until nothing's touching anymore, which then (depending on the font) tends to cause large white holes.
And those CALTs up there could help with similar situations involving "r"-induced crashes, keeping eager mouse cursors away from InDesign's kerning/tracking palette. If people find the CALT feature, that is.

I like your "lateral" idea too though, Craig.

Chris: nice! I don't assume Eben's presentation is documented anywhere? Wish I could see that.

guifa's picture

eliason: interesting, I actually had the same effect but only on the last one.

Nina, I tried livetracing it and playing around but I didn't get very far. My two other thoughts were to finish the x/w/v/y a little lower and pull the tip of the r up and a little to the right, or to start it lower and bring it further to try to fill in the space.... I didn't get very far though (note the file name) but maybe just to give a visual idea. Top right is the original

[edit: it seems my attempt was suggested by eliason while I was working up the image, dang I'm slow]

dezcom's picture

Email Eben or maybe he will see your post.


nina's picture

Heh, those look a bit sad, Matthew. "Bow down before the almighty 'y'!" :-)
Although you might be on to something in terms of disambiguation, say in "rn" vs "m" combos. Hmm.
I guess in the display cut, I could pull the "r"'s arm up over the x-line too. A bit like in Eloquent? Although I'm not convinced that reads well at all.

ebensorkin's picture

Nina wrote to me (thanks!) so I am going to reply here:

I have been planning on posting something but MA at Reading is keeping me very busy and has changed my ideas a bit as well.

The main idea I would like to put across is that for a text design for both technical and typographic reasons 99% of the time it is going to be a bad idea to design a font so that it completely relies on special shapes and scripting. The same can be said about kerning. A design should be acceptable without kerning. Having designed and spaced things really well you can think about kerning. Having done the kerning you may actually be done. If not you may want to look at special shapes and CALT. But these shapes need to solve problems not achievable in some other way.

However if your design is such that some combinations still seem overly awkward or just not as good as they could be - perhaps you have a wide f for reasons of style and want to support several languages- you may want to look at special contextual shapes.

Let me address some of the combinations specifically: to begin with rn should simply work. If ro looks a little loose then so be it. CALT isn't always available. A good solution for rn should be your default. Refinements could be made for a r+o r+e and maybe another for r at the end of a word and or with some punctuation (eg r+. but not r+;).

About the r in this thread: pull it back the arm of the r rather than bending it. I don't think the change to the y is working.

Another example is f+ other letters. Combinations like fi , fl, fh are classics, but even tougher is få fà fā fä and then tougher still is fï fî and then perhaps the very worst is fì.

But the design and its purpose and then fitting of that design that should drive your choice rather than filling a list of possible features to add. Whatever you do with special shapes should not be noticeable to a reader but should instead solve things so that a clash or gap that would otherwise be problem is avoided and reading is made easier.

One of these days I will get round to making a list available but I after I have worked on them and can comment.

It is worth remembering that good letter design, spacing, and even good document design are all more important than this.

There is more to say but I need to sleep!

cerulean's picture

I think the pairs you're highlighting ought to be spaced significantly further apart, with or without the alternate shapes. "rt" in particular seems like you're sacrificing some legibility for the sake of "even color". I do like how the alternates work, though, and I think they'll work even better when you give them some room.

ebensorkin's picture

I want to echo what Kevin is saying - the examples shown do seem mashed together. The combos rv rw ry will have a degree of gappiness even when you have done your best with them. In order to let the shapes be clear you need a space between r and the other letter. The rt is too tight.

Also, the w y & v can be designed so they are more narrow and/or they can have curved strokes if you think the gap is a problem.

nina's picture

Wow, thanks for the input.

Re "relying" on the special effects ;-), honestly I don't think the crash in my example is horrid if it happens (like for Word users or whatever). I'd rather have the crash than a much-larger gap. That said, I think it's nice to be able to avoid the crash, sort of as a bonus. Dunno if that makes it gimmicky; but to be honest I don't quite get the purism. Should it really be better to generally make the spacing "too loose" just so the people without OT support won't see crashes? That strikes me as a more invasive measure that would also influence the nature of the design a lot more (as well, of course, as having differently shaped v/w/y glyphs).

That said, thanks guys for pointing out the too-tight "rt". I'm about to sort-of finalize the lc spacing, but that one I hadn't really seen. (I know, suboptimal timing.)

"About the r in this thread: pull it back the arm of the r rather than bending it. I don’t think the change to the y is working."
Oh. If you have the time, I'd of course be very interested to hear why it's not working for you.
FWIW, my main thought behind trying it this way round was that I don't want the "r" to lose any of its "r"-ness (which is kind of a fragile concept to begin with) by making it narrower; but on the other hand I don't think a "y" stops being readable as a "y" if it's missing a serif. So the idea is remove the clutter/noise that doesn't carry meaning, to make the bit that does carry meaning stand out more and give them more room. If that makes any sense.

"But the design and its purpose and then fitting of that design that should drive your choice rather than filling a list of possible features to add."
*nod* That's also how this came along. I've been impressed with how much "feed-back" there has been from spacing back to the design of individual glyphs, like realizing some glyphs were very slightly too wide or something. With the "r" though, as I said above, I'd like it to keep its current width but am exploring possibilities to make it fit with the others. (Although I'm now thinking that I will probably space its right side a hair more loosely overall.)
Cheers :)

Bendy's picture

I like the idea of this. The printout above looks cool. It looks like r needs more space but hard to tell without seeing it next to other letters that don't crash so much. But I guess it makes sense to have this as a selectable feature (not sure which though).

Did you see this thread about f with umlauts...there's a clever animation by Jos (scroll to 3 Jan 2008) to show how kern, liga and calt features can combine to stop awkward crashes. I was following a similar idea when I came to my f ligs in Mint (not yet posted) and ended up with several different widths for the top of f.

nina's picture

Yeah, I ended up spacing the "r" quite a bit more loosely. Maybe still too tight, though…

New variants with "k" and lslash (partly counterbalanced with kerning), any special opinions on these?

Polish ("wstrząsnęły", what a great test word):

Danish ("beskyttelse", yes I know that "k" is horridly loose on the left side):

(For those into blurring things to simulate parafoveal action, note how in the very bottom-right there is a little bright notch between the "k" and "y", which I would guess can only help – but what do I know!)

Maybe I should have put this into the Critique section.


Thanks for pointing out that other thread, Ben – I had seen and bookmarked it before but somehow forgot about its existence again. There's just too much good reading round here! :)

Bendy's picture

Looks like you're testing this *very* thoroughly! Good on you ;)
I wonder if ky could do with a smidge more space. Look as besky; the first three letters are quite further apart I think. I can't imagine how much work it turns out to be to properly space a font.

nina's picture

Good question – I find that "besky" almost impossible to judge because the "sk" is so horridly off. :-\
Also, I must say I find it really hard to judge the spacing for glyphs that have sideways open counters, like the "k" (and I guess also the "r" to a degree), in terms of: how much space belongs to the glyph shape and *needs* to be there for the glyph to work; and how much is really the space *between* this glyph and the next, which can be eliminated? (I'm not expecting answers on that one ;-) )

BTW, these printouts are quite a bit too bold, probably. A preliminary specimen booklet is being (digitally) printed at the moment, for my presentation at the end of this week. I'm pretty worried what that will look like…

Randy's picture

I was exploring this back in 2006:

I was messaging with Paul Hunt about opentype this afternoon and we stumbled on a further idea together. What about a semi-serif opentype face, that contextually places serifs based on neighboring glyphs?

I can't remember what happened to this! The goal was a little different. I was aiming to minimize the number of serifs to approach sans typography, but still maintain the *essential* serifs for reading. Figuring out *essential* proves to be tricky!

nina's picture

Wow, that's seriously exciting stuff, Randy. Did you take it any further?

Randy's picture

Uh, yes a bit, but looking back on it now there are things to change. Here is a screenshot of a very simple comparison (sorry for the big image):

Middle column has the contextual subs. Mainly, the subs get rid of one serif in situations where there are serifs on adjacent letters. The sub that is relevant to your thread is a situation like rn and ry where the serif is missing on the second letter.

Note, these letters are not all the same with serifs added or cut off. Sidebarings have been adjusted as well as glyph widths (adjusting internal counters) to harmonize the different types of glyphs, and yet, there is no kerning :-) The smallcaps are just for fun, and looking at them now, seem very similar to Sauna. Anyhoo, that is the end of the road on my hardrive. -- This would be a cool thesis project I think, were I in a program that is. Clearly there is a lot more that could be done.

nina's picture

Thanks for showing this here! I find it hugely exciting actually, and promising.
I'm especially intrigued by what happens in sequences like "rn", or similarly the omission of the x-line serif of the "i" in "ri" – which I'd guess would never be consciously noticed by average readers, but subtly disambiguates it so well from "n".
So this approach doesn't just look good (less busy/noisy than the full semi, and more "stable" in a way than the sans), but actually helps! Yum.
I guess the downside is a font like this would be quite an opus magnum to make… and tricky.

dezcom's picture

You can use classes to reduce the code lines. I do think it is a good exercise in learning feature code as well.


Randy's picture

I guess the downside is a font like this would be quite an opus magnum to make… and tricky.

I do think it is a good exercise in learning feature code as well.

Yes, and yes. As for the rn rm combos, it might make sense to add the right-facing serif on the bottom of the r, but also to firm up the word shape at the baseline. Otherwise in a word like "storm" you get three sticks in a row before the closing serif on the m.

This alludes to some of the trickiness. I'm using my gut about what is needed, and I'm not sure I can trust my gut. It seems to be taking advantage of me in non-type matters! That said, my take is that at serifs help define word shapes by emphasizing its edges. And they also help define the counter space between the letters and inside the letters. That was my premise.

nina's picture

Yeah, when I said "tricky" I was more thinking about figuring out the exact concept (what to sub against what, when) than getting it to work technically (although I'm not saying the latter wouldn't be tricky too!). I wouldn't know how to go about it other than by gut feeling, either.

Randy, out of curiosity, have you ever shown this (middle) sample to laymen?
It'd be interesting to know if they notice anything funny going on at all.

BTW, when I presented my font at my type design class a few days ago, I mostly got incredulous and slightly amused reactions to the contextual subs… I like them though and will keep them in, unless/until someone can convince me they don't help.

dezcom's picture

"...I mostly got incredulous and slightly amused reactions to the contextual subs…"

From your class mates? Young people? That is sad. At that age my fellow students would have been thrilled to see someone push the envelope. I hope they appreciate you underneath it all.


nina's picture

Well they weren't like making fun of it; it was more like the "OMG, I didn't even want to look that deep into that wormhole" kind of reaction. Not everyone's equally obsessed I guess… and I did get a bunch of good feedback about my font from them too, no hard feelings at all. I don't blame them; but I do wish we'd heard anything of substance about readability in this class; if I hadn't been hanging around here, I would at most be hazily aware there might be some differences between making text and display fonts.  :-\

Bendy's picture

>I like them though and will keep them in

I'm glad to hear it. I think it's a really interesting solution you've come up with. I agree with Chris, it's great to be experimenting. And I actually do think they work too.

dezcom's picture

Welcome to the Typophile Academy of Type Design. ;-)


nina's picture

Thanks – and seriously! There should be t-shirts. :-)

nina's picture

So to sum this up (or better: to give an update), this is how I ended up doing the "rt" – I shortened "t"'s crossbar a lot more, maybe too much, or maybe it's too loose now:

In the r + diagonals, I left the design as above, couldn't get it to work with a shortened serif – I might try that again though, once I get the spacing to really be waterproof.
And the f + ascender combos look like this (probably also too tight?)

(These are photos from a digital print at 8pt.)

Randy's picture

Aren't you supposed to be on vacation or something?

Randy, out of curiosity, have you ever shown this (middle) sample to laymen?
It’d be interesting to know if they notice anything funny going on at all.

Nope, this is the first time seen in public. I did show my wife, and she thought it was nice, but would be nicer if it would make a bazillion dollars.

Randy's picture

BTW spacing looks better in the prints than the screen grab. Also, fö looks ok, but that fb combo is too tight. Either needs a lig, a narrow f alt, or a contextual f that joins to b h k etc. This is a very nice typeface. Congratulations.

nina's picture

Thank you, Randy! That means a lot.
I'd like to leave the "fb" (etc) unconnected, specifically for German where grammar rules often forbid the use of connecting ligs, which of course doesn't keep the "f" from crashing into ascender serifs – which I therefore removed. But then I fell into the "whee, now I can space it tighter" trap, which probably nullifies the benefit of it all.
And yeah, I said I wouldn't be working on this ATM. On top of that, I should actually be sleeping now… it's addictive stuff. :)

ebensorkin's picture

You could also just space/kern it more widely. I think joins tend to be harder to make work well than letting in some extra space to keep things clear. Joins are brittle in a way that space is not.

Randy - more space around your lc el might be good if this is for text. Look at the density opf rolling vs second. The space between r & n in morning with the calt seems wide.

It's great fun to see you doing this stuff.

William Berkson's picture

I'm afraid you might be going to a more exotic solution, when a simpler one needs to be there, or at least needs to be there before the exotic one is applied as well. To me your r should look be drawn and spaced to look decent between two n's and two o's. In other words, the r needs to be far enough from the n. If this isn't your starting point, then you're making a lot of trouble for yourself to try to make up for it with kerning and tricks.

After that, for vw you can do what you like as far as reducing the outward extension--pretty common--or with kerning. If these moves don't suit you, then the special alternates might be ticket, as you propose.

dezcom's picture

"On top of that, I should actually be sleeping now… "

Why don't you just work on the Z and z? Making zzz's is almost like sleeping :-)


nina's picture

Thanks for the heads-up, William. I spent a bunch of time trying to space it right – but I'm still very much learning to see… I just looked at it again between "n"s and "o"s and indeed it's probably a bit tight on the right side. So I guess I'll go fiddle with that first!

eliason's picture

Randy, what's the logic of the 'r''s in "warrior" have different bottom treatments? Isn't the shape of the left side of the following letter ('r' and 'i') identical?

William Berkson's picture

>I’m still very much learning to see

I'm a few years more at this than you, but only a few, and I completely understand your frustrations, as I have had the same feelings for a long time.

But I've come to the conclusion that the problem is not me, it's the alphabet :)

By that I mean, there are some letters who design just does not permit the same kind of evenness of color across the word as the other letters. Almost all of the lower case letters are either narrow (i,l) or have a design that permits balancing in "four corners" in the x-height region. Those letters that can't be balanced in the same way are: r v w y. You can "cheat" the r by making it narrow, but if you go too far it also looses it's "r'ness".

So I think that the same ideal of even color in the word just can't apply fully to these four letters as they are going to create "holes" of white space toward the bottom of the x-height area whatever you do. And so we have to go to a lesser standard. That's where I think Jost Hochuli's metaphor in Detail in Typography is useful. He says that 'light from above' is more effective than below. Another way of saying this is that on the lower case we give priority to the x-height spacing over the base line spacing.

So after torturing myself over this stuff I have got to a "don't worry, be happy" place, where I am just looking to get the spacing to look right in the x height area, and not worrying about the lower "holes", other than by having the characters not too wide.

In upper case you also additionally have problems with A F L T P. --Which if you add V,W,Y means that 8 letters, a third of the UC have unsolvable "problems!" The Carolingian Minuscule 'solved' those design flaws in those five Imperial Caps. But my feeling is that the Imperial Caps were never used in text anyway, but were display, used on signs and stone carving, and were wide and widely spaced to give priority to elegance of letter form. But acknowledging the inherent problems is to me reassuring because it explains why everyone who is sensitive to these issues feels a need to kern caps in display.

nina's picture

Thanks William. Truth be told, they're not frustrations. This is just me going down the rabbit hole – which keeps getting deeper and darker. Learning to design type has been an engaging exploration to strange and wonderful places, with rare animals along the way, and Typophiles as great travel guides. And sometimes one needs a machete to clear some serifs out of the way!
Come to think of it, Hrant's jungle metaphor was probably pretty good. :-)
Seriously, of course I've made certain steps (like trying to space that "r") five times instead of once. But all of that helps me to learn and to see, which really is the point of the exercise (well apart from making this font :-) ).

It's an interesting angle that the Latin alphabet actively necessitates the presence of white holes. (It would be interesting to know if / how much these are even *helpful* for readability!) Makes me think of my first typography teacher at design school, who used to remind us to "learn to live with our God-given holes" (that of course never failed to crack us up).
It hasn't occurred to me before to connect that to Hochuli's light thing – interesting. (Another way to look at it would probably be that since the upper half of the x-height is more important in reading than the lower half, that might reinforce the importance of good spacing at the x-line, rather than the baseline.)

However one sees the theories though, I suspect the most difficult step involved is learning to *see* the right amount of space, and I'm definitely not quite there yet. But I am beginning to think that cerulean was probably right when he said somewhere up there that I'm caring more for "even color" than I should.

The rabbit hole continues…

dezcom's picture

"Learning to see is harder than learning to look"


thranduil's picture

Pardon me for hijacking, but ah, Chris, how true. It's frustrating, but the rewarding feeling you get makes up for all those frustrations.

Randy's picture

Eliason: I don't remember!

ebensorkin's picture

I’m caring more for “even color” than I should

I feel that way as well. I am willing to trade clarity for even color more & more. But I still like to have my cake & eat it too when possible.

dezcom's picture

You have to redefine "even color" for yourself. Even color has to take into account the script and language to the extent that you are only lessening jarring to the reader without creating a homogenous blend of sameness. Think of even color as a movement in a symphony--it has life and speaks to you with variety and dynamics. Excessive even color is more like Muzak--the mayonnaise of elevator music that barely distinguishes a melody.


nina's picture

"Think of even color as a movement in a symphony"

Interesting angle!
Sounds like this is something that is greatly helped by experience? I must say I can't really hear the music of my spacing all that well yet.
But I think I've been too afraid of "spottiness". Think I'll try to be a bit more daring :-)

PS: I see what you mean re the language – I set a lot of different languages in this font and it was impressive to see how much the color/"pattern" differs. In Icelandic my font looks darker, and in Greenlandic it just looks incredibly wide. That's pretty exciting stuff. Not only does my font speak languages I can't read, it also alters its voice :)

William Berkson's picture

I would put the musical analogy of even color differently. I think even color is more like keeping a regular beat. Then the varied rhythm and melody play against that underlying regularity. I think 'spottiness' is a distraction, just as a drummer who can't keep the beat steady is, or player putting an extra beat in a measure would be.

The "underlying beat" in type is illustrated by those Fourier transform pictures of text.

Even color and an interesting dance of shapes are in tension, but not opposed. Great typefaces have both.

This is not to say that the alphabet is ever going to have perfect evenness of color. But if you don't get *relatively* even color the "word image" or "bouma" doesn't come together properly. As Mario Feliciano said to me, "You want beautiful words, not beautiful letters." Relatively even color and rhythm is a big part of that.

So, Nina, I would say your concern for even color is very healthy for producing good type, though it's only one of several dimensions of type design.

nina's picture

Now I'm confused as to what even color actually *means*…
Is it not directly opposed to spottiness?
Where is the fine line between "spottiness" and the gaps *necessitated* by the structure of the alphabet, as discussed above?
Is there such a thing as *too* even color? If yes: how much is enough?

How do you "see" it to find the right amount? I've tried blurring, but blurring makes me see a lot of splotches that I think are needed/unavoidable (like "e"s being busier than "o"s and such). I've tried looking at stuff from far away, which might be slightly better. But how do I know it's "good"?

Oh and, if anyone has a pointer to a good/comprehensible source to learn about those Fourier transforms, that would be much appreciated. This is about the twentieth time in my life that I wish I understood them. :-\

dezcom's picture

"Now I’m confused as to what even color actually *means*"

This is the dilemma we all face. Though some of us may be more sure than others what it means to us individually, we should all share a healthy respect for the ambiguity that really exists there for all of us collectively. There is much to be said for the line, "neither a borrower nor a lender be" as a way to TRY to see. What I mean is, don't take away from one glyph more than it takes to be itself and don't give to another glyph more than would secure its identity.

Personally, I can't see a use for blurring or distortion techniques. I would rather set some text then put it away for a while (perhaps long enough to not remember why) and just read it on a new day with as naive an eye as you can muster. You pass the test if you read through it without stopping to bitch at yourself for the pattern :-) Also do this same thing for other type that is not your own--even quite revered type of the admired masters of the craft. You may find a surprise or two.


Syndicate content Syndicate content